US9608942B1 - Digital data processing methods and apparatus for the automated generation of personalized digital content - Google Patents
Digital data processing methods and apparatus for the automated generation of personalized digital content Download PDFInfo
- Publication number
- US9608942B1 US9608942B1 US14/177,618 US201414177618A US9608942B1 US 9608942 B1 US9608942 B1 US 9608942B1 US 201414177618 A US201414177618 A US 201414177618A US 9608942 B1 US9608942 B1 US 9608942B1
- Authority
- US
- United States
- Prior art keywords
- digital data
- recipients
- data processing
- recipient
- digital content
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active, expires
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0277—Online advertisement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0254—Targeted advertisements based on statistics
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/04—Real-time or near real-time messaging, e.g. instant messaging [IM]
-
- H04L51/16—
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/21—Monitoring or handling of messages
- H04L51/216—Handling conversation history, e.g. grouping of messages in sessions or threads
Definitions
- the invention relates to digital data processing and more particularly, by way of non-limiting example, to the automated generation of content that is customized to individual reader behaviors.
- the invention has application, by way of non-limiting example, in the automated generation of customized emails, e.g., as part of marketing and sales email campaigns.
- Email campaigns are the bread and butter of modern digital marketing and are often the most important element of the marketing mix of companies. According to the research firm Forrester, email may drive more than 30% of revenues from repeat customers for retail/e-commerce companies.
- email has its own idiosyncratic characteristics.
- the way users interact with email is different from the way they interact with a physical store or newspaper and different from the way they interact with a website.
- an object of the invention is to provide improved methods and apparatus for digital data processing and more particularly, by way of non-limiting example, for the automated generation of customized digital content.
- a further object of the invention is to provide such methods and apparatus as are suited for the automated generation of customized emails, e.g., as part of marketing and sales email campaigns.
- a further related aspect of the invention is to provide such methods and apparatus as are suited for the generation of emails that are personalized for each recipient.
- digital data processing methods of generating digital content pieces e.g., email messages or portions thereof
- Such methods include the step of generating and digitally transmitting to a digital data device of a recipient a digital content piece that (i) has a call to action to which the recipient can respond and (ii) has a plurality of features selected so as to maximize a probability, P(b 1 ,b 2 , . . . ,b M ,x 1 ,x 2 , . . .
- Further related aspects of the invention provide methods, e.g., as described above, that include generating and transmitting substantially concurrently to digital data devices of a plurality of recipients digital content pieces that each (i) has a call to action to which the respective recipient can respond and (ii) has a plurality of features selected so as to maximize a probability, P(b 1 ,b 2 , . . . ,b M ,x 1 ,x 2 , . . . ,x M ), that the recipient will respond to that call to action, where that probability is defined by the aforementioned relation, and where the digital content pieces generated and transmitted to at least two of the recipients differ from one another.
- Another related aspects of the invention provide methods, e.g., as described above, that include generating and transmitting such digital data content pieces substantially concurrently to a first set of over 100 recipients. Still other related aspects of the invention provide methods, e.g., as described above, where the first set comprises over 1000 recipients; over 10,000 recipients; and/or over 100,000 recipients.
- Other related aspects of the invention provide methods, e.g., as described above, that include valuing the coefficients b 1 ,b 2 , . . . ,b M as a function of a dataset representing a history of actual response to calls to action in a plurality, N, of digital content pieces that have one or more of the aforesaid features previously transmitted to digital data devices of a plurality of second set of respective recipients.
- first set and second sets overlap one another. In still other related aspects of the invention, any of the first and the second sets is a subset of the other set. In still yet other related aspects, the first and second sets do not overlap.
- Further related aspects of the invention provide methods, e.g., as described above, that include valuing the coefficients b 1 ,b 2 , . . . ,b M so as to maximize a likelihood, L(b 1 ,b 2 , . . . ,b M ,data), that predicted responses to calls to action for digital content pieces represented in the dataset match the actual responses represented in the dataset.
- Yet still further aspects of the invention provide methods, e.g., as described above, that include valuing the coefficients, b 1 ,b 2 , . . . ,b M , to maximize a log of the aforesaid likelihood, L(b 1 ,b 2 , . . . ,b M ,data).
- FIGS. 1A-1B depict a digital data processing system according to one practice of the invention.
- FIG. 2 depicts phases executed by a server-based application in a system according to the invention to build and deploy personalized click-through models for digital content, such as emails.
- FIGS. 1A-1B depict a digital data processing system 10 according to one practice of the invention for personalizing the automated generation of digital content.
- That content can constitute email messages and, indeed, the generation of such by companies is focus of the embodiment shown in the drawing and discussed below.
- the techniques and apparatus described here can be applied to the personalization of other digital content that is associated with calls to action to which recipients can respond, e.g., text messages, multimedia (MMS) messages, and the like.
- this may be digital content (e.g., email) generated by entities other than companies, e.g., individuals, governmental organizations, and so forth.
- the illustrated system 10 includes a server digital data device 12 that is coupled via network 14 for communication with client digital data devices 16 - 24 .
- the digital data devices 12 and 16 - 24 comprise conventional desktop computers, workstations, minicomputers, laptop computers, tablet computers, PDAs or other digital data devices of the type that are commercially available in the marketplace, all as adapted in accord with the teachings hereof.
- Each comprises a central processing (CPU), memory (RAM), and input/output (IO) subsections that are of the type conventional in the art and that are configured in accord with the teachings hereof to execute an application 30 (in the case of device 12 ) and applications 32 (in the case of devices 16 - 24 ), all programmed utilizing conventional software engineering techniques to perform the functions attributed herein to those applications and, more generally, to the respective devices 12 and 16 - 24 herein.
- the devices 12 , 16 - 24 may be of the same type, though, more typically, constitute a heterogeneous set of devices.
- Devices 12 and 16 - 24 are configured to execute operating systems, email, text messaging, web browser and/or other applications and functions (regardless of whether software-based) all of the conventional type known in the art as adapted in accord with the teachings hereof.
- Applications 30 and 32 cooperate therewith (e.g., with the email, text message, web browser and other functionality executing on the respective devices 12 and 16 - 24 ) to exchange digital content—and, specifically, for example, in the illustrated embodiment, email messages—and responses to calls to action contained therein, all consistent with the teachings hereof.
- server digital data device 12 Although only a single server digital data device 12 is depicted and described here, it will be appreciated that other embodiments may utilize a greater number of these devices, homogeneous, heterogeneous or otherwise, networked or otherwise, to perform the functions ascribed hereto to application 30 and/or digital data processor 12 . Likewise, although several client digital data devices 16 - 24 are shown, it will be appreciated that other embodiments may utilize a greater or lesser number of these devices, homogeneous, heterogeneous or otherwise, running applications 32 that are, themselves, homogeneous, heterogeneous or otherwise.
- Network 14 comprises one or more networks suitable for supporting communications between server 12 and data devices 16 - 24 .
- the network comprises one or more arrangements of the type known in the art, e.g., local area networks (LANs), wide area networks (WANs), metropolitan area networks (MANs), and or Internet(s).
- LANs local area networks
- WANs wide area networks
- MANs metropolitan area networks
- Internet(s) the Internet
- email has its own, idiosyncratic characteristics.
- the way users interact with email is different from the way they interact with, for example, a website.
- they have intent i.e., they have certain goals in mind. For example, if they come to the site with the goal of buying a sweater, they will search for sweaters, click on sweater product images etc.
- the inventors have realized that the factors that drive clickability of a piece of email content (and, more generally, that drive responses to calls to action of digital content pieces) may include, inter alia,
- an effective approach to personalizing the automated generation of email can incorporate data on every customer (their historical purchases or readings, web site visits, email behavior and so on), the various aspects of the contents available for inclusion in the email, and identify the particular combination of factors for each individual that would make him/her click through the email. In short, the approach should be personalized.
- Such an approach can, for example, include the generation of digital content pieces that are customized in accord with individual recipient behaviors, where each such digital content piece (i) has a call to action to which the recipient can respond and (ii) has a plurality, M, of features selected so as to maximize a probability that the recipient will respond to that call to action, which features take into account that recipient's characteristics, including at least that recipient's responses to one or more calls to action of one or more prior digital content pieces.
- a digital content piece is an email message or portion thereof that includes content such as an image(s), video(s) and/or text and that is associated with a call to action.
- a single email message may include more than one content piece.
- a content piece may comprise the whole or portion of a message generated and transmitted via another protocol and/or medium, e.g., an MMS message, etc.
- a call to action is a means by which the recipient of that content piece can respond to indicate her interest in the piece, e.g., vis-a-vis purchase, obtaining more information, “liking” or other commenting, and so forth.
- a call to action is a button or a link (e.g., labeled “buy now,” “reserve yours today,” “tell me more,” etc.) contained in the same email message as the content piece and, indeed, may be embedded in the piece itself (e.g., as in the case of a link embedded in a content image).
- a call to action can, instead or in addition, be a reply-to email address that the user can effectuate by replying to the email using menu or other options in an email client; a word, phrase or code that the user can type upon visiting a specified web site, and so forth; all, by way of non-limiting example.
- the content piece and call to action may be contained, for example, in an email message
- the response effected by the recipient e.g., by clicking the call to action
- a recipient's action vis-a-vis a call to action may result in opening a web browser (e.g., effected via a hidden or obscure code traceable to that email message and/or the recipient thereof).
- a content piece may have one or more of the following features, preferably, 5 or more of those features, still more preferably, 10 or more of those features and, yet still more preferably, 15 or more of those features:
- each news-article oriented content piece may have one or more of the following features, preferably, 3 or more of those features, still more preferably, 6 or more of those features and, yet still more preferably, 9 or more of those features:
- each user-product combination content piece may have one or more of the following features, preferably, 3 or more of those features, still more preferably, 6 or more of those features and, yet still more preferably, 9 or more of those features:
- each user-news-article-headline combination content piece may have one or more of the following features, preferably, 3 or more of those features, still more preferably, 6 or more of those features and, yet still more preferably, 9 or more of those features:
- Illustrated digital data device 12 and, more particularly, application 30 generates personalized click-through models for emails so as to maximize click-through rates, e.g., on email campaigns that can also be generated by that application.
- the application 30 executes a three step process to build and deploy personalized click-through models for emails. It executes a fourth phase to generate and transmit emails based on that model.
- Data collection phase Run email campaigns over a defined time period; include (randomly chosen) content pieces in emails. Collect data on which recipients open and click on what content pieces. Alternatively, or in addition, such data can be discerned from prior email campaigns. See, Step 40 .
- Feature engineering phase Extract predictive variables (aka features) on email recipients, the various aspects of the content pieces, and their interactions thereof. See, Step 42 .
- Model development phase Using the data collected, create and calibrate click-prediction models. See, Step 44 .
- Email campaign phase Using model created/calibrated in the model development phase, generate and transmit to customers digital content pieces that maximize the likelihood of customer response. See step 45 .
- the way users interact with emails can be uniquely different from the way they interact with websites, stores, newspapers etc.
- the application 30 collects data on users actual interactions with emails. Past user data (purchases, users interactions with the website etc) are necessary but not sufficient to build an accurate model.
- the application 30 compiles or otherwise obtains a history of actual responses to calls to action contained in digital content pieces sent to a first set of customers, potential customers or other recipients representative of same (collectively, “customers” or “recipients”). This can be done by randomly or otherwise sampling records from prior email campaigns, e.g., to identify emails transmitting a sufficiently wide sampling of content pieces and associated calls to action to customers and to identify the customer responses thereto. More typically, however, this is done by effecting a “data collection” email campaign using application 30 to generate such emails and to track recipient responses.
- the transmission of such content pieces from server 12 to the client digital data devices 16 - 20 of a first set of customers is graphically depicted by emails icons 46 .
- the customer responses are graphically depicted by arrows 48 .
- X's on those arrows indicate responses not made—i.e., calls to action to which a customer did not respond.
- three content pieces 46 (each, for sake of simplicity, comprising a separate email message) and their associated calls to action are shown as having been transmitted to client device 16
- the customer who received those pieces on that device only responded to calls to action in only two of them.
- the drawing shows no responses to calls to action contained in the four digital content pieces 46 transmitted to device 20 , and only one response to calls to action contained associated with the two content pieces 46 transmitted to device 18 .
- Application 32 of the client digital data devices comprises email clients or other functionality that enable the respective users of those devices to view digital content pieces received by the respective devices and, if desired, to respond to calls to action associated therewith.
- the application 30 extracts a dataset from the history of emails and customer responses developed in the data collection phase. To that end, for each of the N content pieces viewed by its respective recipient (e.g., as evidenced by opening of the email message in which it was contained), the application 30 discerns an enumeration of the M digital content piece/recipient feature combinations (where those features are comprise one or more of the features described in the section “Content Pieces, Features and Calls to Action,” above). This can be done by parsing the digital content piece and/or by obtaining information about the piece and/or the recipient to which it was sent from a database or other records, e.g., in storage associated with device 30 or otherwise. The application 30 captures those per-piece features in an array or other data structure that can be expressed, for example, as follows:
- the application 30 has collected and identified the features in the history of digital content pieces and resultant customer responses, as discussed above.
- the application 30 first, defines a target variable, y i , for each content piece. Specifically, for every customer/content-piece/email combination, if the customer opened the email but did not click on the content piece, the application 30 sets the target variable, y i , for that content piece to zero. If the user clicked on the content piece, the application 30 sets the target variable, y i , to one.
- b j t+1 b j t ⁇ ( y t ⁇ P ( b 1 t ,b 2 t , . . . ,b N t ,x t1 ,x t2 , . . . ,x tM )) x t,j
- the dependent (target) variable For a given email recipient, let us define the dependent (target) variable as y which is equal to 0 for opened but not clicked emails, and is equal to 1 for opened and clicked emails. Let's also define the independent variables (aka predictors, features) as a vector x j where j corresponds to index of a given independent variable and can vary from 1 to M, where M is number of independent variables.
- the set of parameters b 1 , . . . ,b M that maximizes the log-likelihood function LL(b 1 ,b 2 , . . . ,b M ,data) can be found by using Stochastic Gradient Descent algorithm described below.
- b j t + 1 b j t - ⁇ ⁇ ⁇ LL ⁇ ( b 1 , b 2 , ... ⁇ , b M , data )
- the application 30 in an email campaign phase, the application 30 generates, for each of one or more customers in a second set of customers, digital content pieces having features that maximize a probability, P(b 1 ,b 2 , . . . ,b M ,x 1 ,x 2 , . . . ,x M ), that the customer will respond to a call to action associated with that piece.
- M characterize the content piece and, rather, because at least one (or, more preferably, some) of them characterize aspect of the customer to whom the content piece is directed, generation of digital content pieces in this manner results in pieces that differ from one another—and, more specifically, where the digital content pieces generated and transmitted to at least two of the recipients differ from one another.
- the application 30 effects digital transmission of those content pieces substantially concurrently (i.e., within the time period typical to that for the generation and transmission of email messages in bulk to a plurality of recipients) to the respective customers via email, with one or more content pieces per email.
- the application generates and transmits such content pieces (e.g., emails) to over 1000 customers at a time; other embodiments, to over 10,000 customers at a time; and, in still other embodiments, to over 100,000 customers at time.
- the generation of content pieces yielding higher rates of responses to associated calls to action is made possible through utilization model-based customization, as discussed herein.
- Applications 32 executing on the digital data devices 16 - 24 permit the respective customers to view those emails/content pieces and to respond to them, e.g., by clicking on or otherwise taking action with respect to the pieces' associated calls to action.
- the transmission of such emails from server 12 to the client digital data devices 16 - 20 as part of such a campaign is graphically depicted by emails icons 46 .
- the customer responses are graphically depicted by arrows 48 .
- no X's are shown on the arrows of FIG. 1B , suggesting that due to maximization of the probabilities discussed above, the recipients in the second set respond to all of the calls to action associated with digital content pieces generated utilizing the model and model coefficients discussed above. In actual practice, this may not the case, though, such maximization is intended to increase the rate of user response over systems that do not employ the teachings hereof.
- the recipients of digital content pieces (e.g., emails) in the data collection phase need not correspond to those of the email campaign phase.
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Probability & Statistics with Applications (AREA)
- Information Transfer Between Computers (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
P(b 1 ,b 2 , . . . ,b M ,x 1 ,x 2 , . . . ,x M)=exp(Σj=1, . . . ,M b j x j)/(1+exp(Σj=1, . . . ,M b j x j))
-
- where
- x1,x2, . . . ,xM are values for each of a plurality, M, of features characterizing the digital content piece and/or the recipient,
- b1,b2, . . . ,bM are respective coefficients for each of the values x1,x2, . . . ,xM.
Description
-
- Product/Offer recommendations (used heavily by retail/e commerce/travel/hospitality companies
- News article recommendations (used heavily by media companies)
P(b 1 ,b 2 , . . . ,b M ,x 1 ,x 2 , . . . ,x M)=exp(Σj=1, . . . ,M b j x j)/(1+exp(Σj=1, . . . ,M b j x j))
-
- where
- x1,x2, . . . ,xM are values for each of a plurality, M, of features characterizing the digital content piece and/or the recipient,
- b1,b2, . . . ,bM are respective coefficients for each of the values x1,x2, . . . ,xM
L(b 1 ,b 2 , . . . ,b M,data)=Πi=1, . . . N;y
-
- where
- Πi=1, . . . N;y
i =1P(b1,b2, . . . ,bM,xi1,xi2, . . . ,xiM) represents a product of probabilities, P(b1,b2, . . . ,bM,x1,x2, . . . ,xM), for digital content pieces having features of those digital content pieces that were (i) previously transmitted to digital data devices of the plurality of respective recipients in the second set, and (ii) that were opened (or otherwise accessed) by their respective recipients, and (iii) to which the respective call to action was responded to by that recipient. - Πi=1, . . . N;y
i =0(1−P(b1,b2, . . . ,bM,xi1,xi2, . . . ,xiM)) represents a product of probabilities, P(b1,b2, . . . ,bM,x1,x2, . . . ,xM), for digital content pieces having features of those that were (i) previously transmitted to digital data devices of the plurality of respective recipients in the second set, and (ii) that were opened (or otherwise accessed) by there respective recipients, and (iii) to which the respective call to action was not responded to by that recipient.
b j t+1 =b j t−η(y t −P(b 1 t ,b 2 t , . . . ,b N t ,x t1 ,x t2 , . . . ,x tM))x t,j
-
- where,
- η is learning rate set by trial and error,
- P(b1 t,b2 t, . . . ,bN t,xt1,xt2, . . . ,xtM)) is the probability, P(b1,b2, . . . ,bM,x1,x2, . . . ,xM), evaluated using the coefficients bj, for j=1 . . . M
-
- the various physical aspects of that content, e.g., black-and-white vs. color image, size of image, presence of model
- the email recipients historical interactions with that company in general and the specific types of content featured in the email
- the interactions between the two sets of factors above (e.g., a user may find black-and-white images of basics like socks eminently clickable, but may not click on black-and-white images of fashion sweaters)
-
- Category
- Brand
- Newness indicator
- Gender indicator
- Special indicators (e.g. Halloween, Easter, Anniversary)
- Keywords in product description (e.g., cashmere, BOGO—Buy One Get One)
- Price range
- Deal range
- All-Black-image indicator
- All-White-image indicator
- Modeled/Unmodeled indicator (if the image is of a model wearing the product or just the product itself)
- Primary color of product image
- Resolution of product image
- Presence of caption under product image
- Number of lines of caption text
- Indicator if price/discount is mentioned in caption
- Indicator if adjacent product is from the same category
- Indicator if adjacent product is 10
- Total number of products featured in the email
- Indicator if product image is visible above the fold for a desktop/laptop screen
- Indicator if product image is visible above the fold for a tablet screen
- Indicator if product image is visible above the fold for a smartphone screen
-
- Number of words in headline
- Color of font used in headline
- Number of lines of caption text
- Indicator if there is accompanying image
- Indicator if accompanying image is black-and-white
- Size of accompanying image
- Total number of news article headlines featured in the email
- Indicator if content is visible above the fold for a desktop/laptop screen
- Indicator if content is visible above the fold for a tablet screen
- Indicator if content is visible above the fold for a smartphone screen
- Actual text of news article headline
-
- Category previously purchased indicator
- Category recency (if purchased)
- Brand previously purchased indicator
- Brand recency (if purchased)
- Product in ecommerce shopping cart indicator
- Category in ecommerce shopping cart indicator
- Brand in ecommerce shopping cart indicator
- Product browsed in past 7 days indicator
- Category browsed in past 7 days indicator
- Brand browsed in past 7 days indicator
- Product clicked-through-in-email indicator
- Category clicked-through-in-email indicator
- Brand clicked-through-in-email indicator
- Bought in that price range before indicator
- Category previously purchased ONLINE indicator
- Product complements customer's previous buy indicator (e.g., Shirt and Tie)
- Affinity score for that user-product combination (this is optional and can come from external sources, if available)
-
- Indicator for whether the user has previously read articles in the topics pertaining to the news article headline
- Indicator for whether the user has previously clicked-through-in-email article headlines pertaining to the topics referenced by the news article headline
- Number of days since previous read of any topic
- Number of days since previous read of relevant topics
- Number of articles read in past 7 days
- Number of articles read in past 30 days
- Number of relevant articles read in past 7 days
- Number of relevant articles read in past 30 days
- Number of article headlines clicked-through-in-email in past 7 days
- Number of article headlines clicked-through-in-email in past 30 days
-
- where
- xi,j for i=1 . . . N and j=1 . . . M, are values for each of the respective M digital content piece/recipient features for the respective digital content pieces N
P(b 1 ,b 2 , . . . ,b M ,x 1 ,x 2 , . . . ,x M)=exp(Σj=1, . . . ,M b j x j)/(1+exp(Σj=1, . . . ,M b j x j))
-
- where,
- x1,x2, . . . ,xM are values for each of the respective M features characterizing the digital content piece and/or recipient thereof,
- b1,b2, . . . ,bM are respective coefficients for each of the values x1,x2, . . . ,xM
L(b 1 ,b 2 , . . . ,b M,data)=Πi=1, . . . N;y
-
- where
- Πi=1, . . . N;y
i =1P(b1,b2, . . . ,bM,xi1,xi2, . . . ,xiM) represents a product of probabilities, P(b1,b2, . . . ,bM,x1,x2, . . . ,xM), for digital content pieces having features of those digital content pieces that were (i) previously transmitted to digital data devices of the plurality of respective recipients of the first set, and (ii) that were opened (or otherwise accessed) by there respective recipients, and (iii) to which the respective call to action was responded to by that recipient. - Πi=1, . . . N;y
i =0(1−P(b1,b2, . . . ,bM,xi1,xi2, . . . ,xiM)) represents a product of probabilities, P(b1,b2, . . . ,bM,x1,x2, . . . ,xM), for digital content pieces having features of those that were (i) previously transmitted to digital data devices of the plurality of respective recipients of the first set, and (ii) that were opened (or otherwise accessed) by there respective recipients, and (iii) to which the respective call to action was not responded to by that recipient.
b j t+1 =b j t−η(y t −P(b 1 t ,b 2 t , . . . ,b N t ,x t1 ,x t2 , . . . ,x tM))x t,j
-
- where,
- η is learning rate set by trial and error,
- P(b1 t,b2 t, . . . ,bN t,xt1,xt2, . . . ,xtM)) is the probability, P(b1,b2, . . . ,bM,x1,x2, . . . ,xM), evaluated using the coefficients bj, for j=1 . . . M
P(b 1 ,b 2 , . . . ,b M ,x 1 ,x 2 , . . . ,x M)=exp(Σj=1, . . . ,M b j x j)/(1+exp(Σj=1, . . . ,M b j x j))
-
- where bj is the model coefficient that corresponds to the independent variable xj
L(b 1 ,b 2 , . . . ,b M,data)=Πi=1, . . . N;y
LL(b 1 ,b 2 , . . . ,b M,data)=Σi=1, . . . N y i log(P(b 1 ,b 2 , . . . ,b M ,x i1 ,x i2 , . . . ,x iM))+(1−y i)log(1−P(b 1 ,b 2 , . . . ,b M ,x i1 ,x i2 , . . . ,x iM))=Σi=1, . . . N(y iΣj=1, . . . ,M b j x i,j−log(1+exp(Σj=1, . . . ,M b j x i,j))
for j=1, . . . ,M
-
- where
- η is learning rate set by trial and error,
Claims (14)
b j t+1 =b j t−η(y t −P(b 1 t ,b 2 t , . . . ,b N t ,x t1 ,x t2 , . . . ,x tM))x t,j
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/177,618 US9608942B1 (en) | 2014-02-11 | 2014-02-11 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
PCT/US2015/014273 WO2015123054A1 (en) | 2014-02-11 | 2015-02-03 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
EP15748468.4A EP3105728A4 (en) | 2014-02-11 | 2015-02-03 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
JP2016551266A JP6588023B2 (en) | 2014-02-11 | 2015-02-03 | Digital data processing method, computer program, and storage medium |
US15/454,550 US10559015B2 (en) | 2014-02-11 | 2017-03-09 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/177,618 US9608942B1 (en) | 2014-02-11 | 2014-02-11 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/454,550 Continuation US10559015B2 (en) | 2014-02-11 | 2017-03-09 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
Publications (1)
Publication Number | Publication Date |
---|---|
US9608942B1 true US9608942B1 (en) | 2017-03-28 |
Family
ID=53800535
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US14/177,618 Active 2034-05-08 US9608942B1 (en) | 2014-02-11 | 2014-02-11 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
US15/454,550 Active US10559015B2 (en) | 2014-02-11 | 2017-03-09 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
Family Applications After (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/454,550 Active US10559015B2 (en) | 2014-02-11 | 2017-03-09 | Digital data processing methods and apparatus for the automated generation of personalized digital content |
Country Status (4)
Country | Link |
---|---|
US (2) | US9608942B1 (en) |
EP (1) | EP3105728A4 (en) |
JP (1) | JP6588023B2 (en) |
WO (1) | WO2015123054A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180032888A1 (en) * | 2016-08-01 | 2018-02-01 | Adobe Systems Incorporated | Predicting A Number of Links an Email Campaign Recipient Will Open |
US10559015B2 (en) | 2014-02-11 | 2020-02-11 | Salesforce.Com, Inc. | Digital data processing methods and apparatus for the automated generation of personalized digital content |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11210709B2 (en) | 2018-06-21 | 2021-12-28 | International Business Machines Corporation | Methods and systems for generating personalized call-to-action elements |
EP4049134A4 (en) * | 2019-10-25 | 2023-12-13 | Capital One Services, LLC | Recipient-side dynamic customization of an email intended for the recipient or pre-dispatch dynamic customization of email instances of an email |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040204975A1 (en) * | 2003-04-14 | 2004-10-14 | Thomas Witting | Predicting marketing campaigns using customer-specific response probabilities and response values |
US20070288304A1 (en) * | 2006-06-08 | 2007-12-13 | Adknowledge, Inc. | System and method for behaviorally targeted electronic communications |
US20080275753A1 (en) * | 2007-05-01 | 2008-11-06 | Traffiq, Inc. | System and method for brokering the sale of internet advertisement inventory as discrete traffic blocks of segmented internet traffic. |
US20090019122A1 (en) * | 2007-07-12 | 2009-01-15 | Fatdoor, Inc. | Direct mailing in a geo-spatial environment |
US20110055000A1 (en) * | 2009-08-27 | 2011-03-03 | Xin Zhang | Predicting email responses |
US20120150626A1 (en) | 2010-12-10 | 2012-06-14 | Zhang Ruofei Bruce | System and Method for Automated Recommendation of Advertisement Targeting Attributes |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011069073A1 (en) * | 2009-12-03 | 2011-06-09 | Comscore, Inc. | Measuring advertising effectiveness without control group |
US8738440B2 (en) * | 2010-06-14 | 2014-05-27 | International Business Machines Corporation | Response attribution valuation |
US9807152B2 (en) | 2012-07-30 | 2017-10-31 | Nec Corporation | Distributed processing device and distributed processing system as well as distributed processing method |
US9608942B1 (en) | 2014-02-11 | 2017-03-28 | CQuotient, Inc. | Digital data processing methods and apparatus for the automated generation of personalized digital content |
-
2014
- 2014-02-11 US US14/177,618 patent/US9608942B1/en active Active
-
2015
- 2015-02-03 WO PCT/US2015/014273 patent/WO2015123054A1/en active Application Filing
- 2015-02-03 EP EP15748468.4A patent/EP3105728A4/en active Pending
- 2015-02-03 JP JP2016551266A patent/JP6588023B2/en active Active
-
2017
- 2017-03-09 US US15/454,550 patent/US10559015B2/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040204975A1 (en) * | 2003-04-14 | 2004-10-14 | Thomas Witting | Predicting marketing campaigns using customer-specific response probabilities and response values |
US20070288304A1 (en) * | 2006-06-08 | 2007-12-13 | Adknowledge, Inc. | System and method for behaviorally targeted electronic communications |
US20080275753A1 (en) * | 2007-05-01 | 2008-11-06 | Traffiq, Inc. | System and method for brokering the sale of internet advertisement inventory as discrete traffic blocks of segmented internet traffic. |
US20090019122A1 (en) * | 2007-07-12 | 2009-01-15 | Fatdoor, Inc. | Direct mailing in a geo-spatial environment |
US20110055000A1 (en) * | 2009-08-27 | 2011-03-03 | Xin Zhang | Predicting email responses |
US20120150626A1 (en) | 2010-12-10 | 2012-06-14 | Zhang Ruofei Bruce | System and Method for Automated Recommendation of Advertisement Targeting Attributes |
Non-Patent Citations (2)
Title |
---|
International Search Report and Written Opinion for Application No. PCT/US2015/014273, mailed Jun. 18, 2015, (11 pages). |
Stochastic Gradient Descent (Logistic Regression and Stochastic Gradient Tranining, Charles Elkan, Jan. 29, 2010, https://web.archive.org/web/20100701234413/http://cseweb.ucsd.edu/˜elkan/250B/logreg.pdf. * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10559015B2 (en) | 2014-02-11 | 2020-02-11 | Salesforce.Com, Inc. | Digital data processing methods and apparatus for the automated generation of personalized digital content |
US20180032888A1 (en) * | 2016-08-01 | 2018-02-01 | Adobe Systems Incorporated | Predicting A Number of Links an Email Campaign Recipient Will Open |
US10997524B2 (en) * | 2016-08-01 | 2021-05-04 | Adobe Inc. | Predicting a number of links an email campaign recipient will open |
Also Published As
Publication number | Publication date |
---|---|
WO2015123054A1 (en) | 2015-08-20 |
JP2017506786A (en) | 2017-03-09 |
US20170186053A1 (en) | 2017-06-29 |
US10559015B2 (en) | 2020-02-11 |
EP3105728A4 (en) | 2017-08-30 |
JP6588023B2 (en) | 2019-10-09 |
EP3105728A1 (en) | 2016-12-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20240062271A1 (en) | Recommendations Based Upon Explicit User Similarity | |
US11798052B2 (en) | Systems, apparatuses, and methods for extracting inventory from unstructured electronic messages | |
US9384502B2 (en) | Techniques for organizing and presenting deal content | |
US8554638B2 (en) | Comparative shopping tool | |
US10559015B2 (en) | Digital data processing methods and apparatus for the automated generation of personalized digital content | |
CN102132308B (en) | Information sharing in an online community | |
TW200849131A (en) | A system and device for social shopping on-line | |
JP2019164706A (en) | Information processing device, information processing method, and program | |
Schaeffer et al. | A framework for informing consumers on the ecological impact of products at point of sale | |
Orimoloye et al. | Differential effects of device modalities and exposure to online reviews on online purchasing: A field study | |
JP2018169987A (en) | Recommendation information providing system and recommendation information providing method | |
GB2567644A (en) | Method and system for supplying commodities | |
US10915938B2 (en) | Including instructions upon item procurement | |
Kamaruddin | Comparative Study on Sentiment Analysis Approach for Online Shopping Review | |
US11475079B2 (en) | System and method for efficient multi stage statistical website indexing | |
Jain et al. | An Investigation into the relationship between AI enabled Chabot Interface and Online buying behavior of Consumers in Delhi NCR Region | |
Slik et al. | Data-driven Direct Marketing via Approximate Dynamic Programming | |
Somya et al. | A novel approach to collect and analyze market customer behavior data on online shop | |
US20230410182A1 (en) | Systems and methods handling mobile technology communications | |
JP7453191B2 (en) | Information processing device, information processing method, and information processing program | |
Patnaik et al. | Business Analytics Using Recommendation Systems | |
US20150332343A1 (en) | Behavioral tracking system and method in support of high-engagement communications | |
US20220101402A1 (en) | System and method for determining an attractiveness value of a product displayed on an online platform, and generating insights respective thereof | |
VidhyaPriya et al. | Connecting social media to eCommerce using microblogging and artificial neural network | |
Kübler | Leveraging Unstructured Data in Channel Attribution Theory |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: CQUOTIENT, INC., MASSACHUSETTS Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:RAMAKRISHNAN, VISHWAMITRA S.;POPKOV, YVEGENIY;REEL/FRAME:033375/0847 Effective date: 20140718 |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
AS | Assignment |
Owner name: SALESFORCE.COM, INC., CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:DEMANDWARE, INC.;REEL/FRAME:052873/0975 Effective date: 20200406 Owner name: SALESFORCE.COM, INC., CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:CQUOTIENT, INC.;REEL/FRAME:052875/0558 Effective date: 20200406 |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 4 |